This award is for continued support for a program of research focusing on nonlinear, econometric methods. The research is based on the premise that an econometric specification is an approximation to an underlying data generating mechanism. In keeping with this perspective, methodologies will be developed that improve the approximation as more information becomes available and permits more reliable inference at each intermediate stage of model evolution. The scientific importance of this research derives from the fact that the goal of empirical research is to make an inference about an economic proposition per se, not about an economic proposition entangled with a potentially misspecified or incorrect model. With standard approaches to specification tests, one cannot be sure that the estimate associated with an inference is accurate. The objective of this research can be viewed as the development of a model selection strategy that guarantees that the final estimate is accurate. This objective is to be accomplished at a level of generality that encompasses most nonlinear, econometric inference procedures. The basic idea is to endow a procedure with nonparametric properties by replacing the structural model with a truncated series expansion, or the error density with a truncated expansion, or both. By letting the truncation point grow adaptively with sample size, ultimate convergence to the underlying data generating mechanism is assured. Applications to conditionally heterogeneous time-series such as occur in finance and monetary and macro economics will be undertaken. Bayesian methods will also be investigated.
